The Journal of Neuroscience, October 31, 2012 • 32(44):15377–15387 • 15377

Development/Plasticity/Repair Sorting within the Marginal Zone via Robo-Mediated Inhibition of N-Cadherin Controls Formation

Nozomi Sakai,1 Ryan Insolera,3,4 Roy V. Sillitoe,1 Song-Hai Shi,3,4 and Zaven Kaprielian1,2 1Dominick P. Purpura Department of Neuroscience and 2Department of Pathology, Albert Einstein College of Medicine, Bronx, New York, 10461; 3Developmental Biology Program, Memorial Sloan-Kettering Cancer Center, New York, New York 10065; and 4Neuroscience Graduate Program, Weill Cornell Medical College, New York, New York 10065

The of spinal projection transmit sensory information to the brain by ascending within highly organized longitudinal tracts. However, the molecular mechanisms that control the sorting of these axons within the spinal cord and their directed growth to poorly defined targets are not understood. Here, we show that an interplay between Robo and the cell adhesion molecule, N-cadherin, sorts spinal commissural axons into appropriate longitudinal tracts within the spinal cord, and thereby facilitates their brain targeting. Specifically, we show that d1 and d2 spinal commissural axons join the within the spinal cord and target the in chick embryos, and that these axons contribute to the spinocerebellar projection in transgenic reporter mice. Disabling Robo signaling or overexpressing N-cadherin on these axons prevents the formation of the lateral funiculus and the spinocerebellar tract, and simulta- neously perturbing Robo and N-cadherin function rescues both phenotypes in chick embryos. Consistent with these observations, disabling Robo function in conditional N-cadherin knock-out mice results in a wild-type-like lateral funiculus. Together, these findings suggest that spinal projection axons must be sorted into distinct longitudinal tracts within the spinal cord proper to project to their brain targets.

Introduction Flanagan, 2007; Sakano, 2010). Accordingly, the spatial arrange- The stereotypical growth of axons to their synaptic targets is a ment of ascending axons within distinct longitudinal funiculi in crucial phase of neural circuit formation. In the vertebrate spinal the spinal cord proper may dictate their long-range projections to cord, ascending projection neurons extend axons into the mar- appropriate brain targets. ginal zone where they are arranged within distinct longitudinal Dorsal spinal commissural neurons are a major class of pro- tracts, based on the locations of their brain targets (Brodal, 1998; jection neurons whose axons cross the midline en route to the Willis, 2007). Topographically organized axonal projections es- brain (Brodal, 1998). Although some evidence suggests that tablish stereotyped patterns of connectivity throughout the ner- Atoh1-derived d1 axons project to the cerebellum (Bermingham vous system and can define the routes that axons follow to reach et al., 2001), the brain targets of genetically distinct classes of their synaptic targets (Garel and Rubenstein, 2004; Luo and dorsal spinal neurons have not been explicitly identified. After crossing the midline, most spinal commissural axons execute a rostral turn and project along a medial longitudinal commissural Received May 9, 2012; revised Aug. 23, 2012; accepted Sept. 3, 2012. (MLc) trajectory adjacent to the floor plate and form the ventral Author contributions: N.S. and Z.K. designed research; N.S., R.I., and R.V.S. performed research; N.S. and S.-H.S. contributed unpublished reagents/analytic tools; N.S. analyzed data; N.S. and Z.K. wrote the paper. funiculus (VF) (Bovolenta and Dodd, 1990; Imondi and Kapri- This work was supported by grants from the National Institutes of Health, National Institute of Neurological elian, 2001; Kadison and Kaprielian, 2004). A subset of these Disorders and Stroke (R01NS038505), and the New Jersey Commission on Spinal Cord Research (CSCR11IRG011) axons, termed intermediate longitudinal commissural (ILc) ax- (Z.K.). We thank the following investigators for constructs: Janice Imai, Tohru Yamamoto, Roger Bradley, Deanna ons, ultimately leave the VF and project in an arcuate manner Benson,andAllanBradley.WealsothankStaceyReeberforherassistancewithmousework.WearegratefultoCarol Mason, Jane Dodd, Jane Johnson, Yimin Zou, Hannes Bu¨low, Jean He´bert, Scott Emmons, Marie Fernandes, Cristina into the lateral funiculus (LF). Aguirre-Chen, and Angela Jevince for providing critical and insightful comments on the manuscript. Anti-mouse The relative positioning of longitudinal axon tracts in the ver- Ncad (6B3) was obtained from the Developmental Studies Hybridoma Bank, developed under the auspices of the tebrate spinal cord is in part controlled by the Robo-Slit guidance National Institute of Child Health and Human Development and maintained by the University of Iowa, Department system. In the chick spinal cord, postcrossing spinal commissural of Biological Sciences (Iowa City, IA). R.V. Sillitoe’s present address: Department of Pathology and Immunology, Baylor College of Medicine, Jan and axons lacking Robo signaling fail to elaborate ILc projections and, Dan Duncan Neurological Research Institute of Texas Children’s Hospital, Houston, TX 77030. instead, extend in hyperfasciculated MLc tracts within the VF The authors declare no competing financial interests. (Reeber et al., 2008). A similar phenotype is observed in the spinal CorrespondenceshouldbeaddressedtoZavenKaprielian,DominickP.PurpuraDepartmentofNeuroscienceand cord of mouse embryos lacking Robo1 and Robo2 or all three Slits Department of Pathology, Albert Einstein College of Medicine, Kennedy Center Room 624, 1410 Pelham Parkway South, Bronx, NY 10461. E-mail: [email protected]. (Long et al., 2004; Jaworski et al., 2010). Nevertheless, the molec- DOI:10.1523/JNEUROSCI.2225-12.2012 ular mechanisms that confine postcrossing commissural axons to Copyright © 2012 the authors 0270-6474/12/3215377-11$15.00/0 the contralateral VF in the absence of Robo function are not 15378 • J. Neurosci., October 31, 2012 • 32(44):15377–15387 Sakai et al. • Robo Inhibits Ncad to Form Spinocerebellar Tract known. The cell adhesion molecule, N-cadherin (Ncad) is re- Histochemistry. All fluorescence immunostaining was performed on quired for the proper positioning of longitudinal axon tracts in 20–50 ␮m cryosections as previously described (Jevince et al., 2006). Drosophila (Iwai et al., 1997), and Robos are capable of attenuat- Primary antibodies used for immunohistochemistry were anti-mouse Ncad (6B3, 1:5; Developmental Studies Hybridoma Bank) and anti- ing Ncad function in a variety of systems (Rhee et al., 2002, 2007; ␤ ␤ Wong et al., 2012). Thus, it is conceivable that Robo-mediated rabbit -galactosidase ( -gal, 1:500; AbD Serotec). Appropriate Al- exa Fluor 488, 568, (1:1000; Invitrogen) or Cy3 (1:200; Jackson inhibition of Ncad-dependent fasciculation regulates funiculi ImmunoResearch) conjugates were used as secondary antibodies. For formation in the developing CNS. 3,3Ј-diaminobenzidine (DAB) photoconversion of fluorescence sig- Here, we use genetic labeling strategies to show that dorsal nals, we stained, as whole mounts, CNS tissues transfected with the d1 spinal d1 and d2 axons project to multiple brain regions includ- or d2 reporter, with anti-rabbit green fluorescent protein (GFP, ing the cerebellum. Moreover, we show that disabling Robo sig- 1:1000; Invitrogen), and subsequently with biotin-conjugated anti- naling or overexpressing Ncad on spinal commissural axons rabbit secondary antibody (1:50; Vector Laboratories). The tissue was prevents these axons from joining the LF, and subsequently dis- further incubated with avidin–biotin–peroxide complex and pro- rupts spinocerebellar tract formation. Importantly, the simulta- cessed for DAB reaction (Vector Laboratories) as previously de- neous perturbation of Robo and Ncad function rescues these scribed (Jevince et al., 2006). AP reaction in whole mounts of CNS axon positioning and targeting phenotypes. Together, our find- transfected with CAG-GPI-AP was performed as previously described (Imondi et al., 2000; Arakawa et al., 2008). For ␤-gal staining, 20 ␮m ings suggest that Robo-mediated inhibition of Ncad is required cryosections from P6 lacZ-reporter mice were fixed in cold 4% para- for sorting spinal projection axons into distinct longitu- formaldehyde for 5 min, washed in 5-bromo-4-chloro-3-indolyl-␤- dinal tracts and that this facilitates their projection to the D-galactopyranoside (X-gal) washing buffer, which contains 2 mM cerebellum. MgCl2, 0.05% sodium deoxycholate, and 0.1% Igepal Ca-630 (Sigma- Aldrich) in PBS, two times for 10 min at 25°C, and incubated in X-gal

reaction solution (1 mg/ml X-gal, 5 mM K3Fe(CN)6 and5mM Materials and Methods ⅐ 3 K4Fe(CN)6 H2O in X-gal washing buffer) at 37°C for 12–18 h. Animals. Fertilized White Leghorn eggs were obtained from Charles River Lineage mapping and retrograde tracing. To carry out fate mapping, we and incubated at 39°C. Ncadflox (Kostetskii et al., 2005), Atoh1CreERT2 (Ma- crossed Rosa26-lacZflox Cre reporter mice with Atoh1CreERT2 or chold and Fishell, 2005), Neurog1CreERT2 (Koundakjian et al., 2007), and Neurog1CreERT2 mice. Tamoxifen (Sigma-Aldrich) in corn oil was ad- Rosa26-lacZflox (Soriano, 1999) mice were obtained from the The Jackson ministered by oral gavage at 2–4 mg per 40 g of 10.5 d pregnant mice. At Laboratory. P4, pups were genotyped for CreER alleles by PCR with following prim- In ovo and in utero electroporation. For in ovo electroporation, plasmid ers: forward 5Ј-GCCTGGTCTGGACACAGTGCC-3Ј and reverse 5Ј-CT construct(s) either alone or together with morpholinos (see below) were GTCTGCCAGGTTGGTCAGTAAGC-3Ј. Genotyping of Rosa26-lacZflox injected into the of E2.5–E3.5 chick neural tubes. Unilateral was performed as described previously (Soriano, 1999). ␤-Gal staining of transfection of the spinal cord was performed by applying five 20–21 V Atoh1CreER;Rosa26-lacZflox mice spinal cord was performed at P6 as pulses for 50 ms at 100 ms intervals (Krull, 2004). In utero electropora- described above. For retrograde tracing of spinocerebellar axons/tracts tion was performed as previously described (Bultje et al., 2009) except originating from Atoh1 progenitor-derived d1 and Neurog1-lineage neu- that plasmids were injected into the central canal of E11–E11.5 mouse rons, we injected 2% Alexa Fluor 555-conjugated wheat germ agglutinin embryos, and five 40 V 50 ms pulses at 950 ms intervals were applied to (WGA) tracer (Invitrogen) in PBS into lobules V–VIII of the cerebellum the spinal cord. Chick and mouse embryos of either sex were used for of P5 Atoh1CreER;Rosa26-lacZflox and Neurog1CreER;Rosa26-lacZflox electroporation. pups. The surgery was conducted as previously described (Reeber et al., Plasmid constructs and morpholinos. The c␤actin-NcadFL construct 2011). For colabeling, 1 d later, at P6, we dissected out the spinal cord and (a gift from R. Bradley), which contains a chick ␤-actin promoter, was performed anti-␤-gal immunostaining of cryosections as described used to overexpress the full-length form of chick Ncad. The following above. plasmids containing a CAG promoter (Niwa et al., 1991) were used for Photodocumentation. Epifluorescence images were captured using a manipulating gene function via electroporation: CAG-humanRobo1⌬C- Nikon Eclipse TE300 microscope. The confocal images were obtained GFP (Hammond et al., 2005; Reeber et al., 2008) and CAG-xenopusNcad⌬E- using a Fluoview 500 microscope (Olympus) and processed with Im- myc (Bozdagi et al., 2004). We used CAG-nlsCre-IRES-GFP for ageJ64 (National Institutes of Health). The images of DAB- or AP- eliminating Ncad expression in Ncadflox mice and CAG-IRES-GFP as the stained CNS whole mounts were acquired on a Zeiss Stemi-2000C stereo corresponding control, and delivered these plasmids into mouse em- dissecting microscope (Carl Zeiss). Brightness and contrast of images bryos via in utero electroporation. CAG-Robo1⌬C-taumCherry was were adjusted using Adobe Photoshop CS. constructed by replacing the GFP sequence in CAG-Robo1⌬C-GFP Quantification and data analysis. For quantification of d2 ILc axons, plasmid (Hammond et al., 2005) with taumCherry, which was ob- we selected a 0.5 mm 2 region in the center of the electroporated area of a tained by PCR from a Neurog1taumCherry construct (Reeber et al., given open-book preparation. All labeled decussated axons that did not 2008). We used Atoh1tauGFP or Atoh1taumCherry, and project parallel to, and in the vicinity of, the floor plate were regarded as Neurog1tauGFP or Neurog1taumCherry plasmids as reporter con- ILc axons. We first normalized d2 commissural axon counts, defined as structs to visualize the cell bodies and axons of d1 and d2 neurons, the number of labeled d2 axons within the floor plate, to the respective respectively (Lumpkin et al., 2003; Nakada et al., 2004; Reeber et al., counts of transfected cell bodies, and verified that the resulting ratios 2008). The Neurog1-reporter construct contains a previously de- were comparable across preparations. The numbers of d2 ILc axons were scribed 0.8 kb dorsal-specific enhancer element, TgN1–13 (Nakada et normalized to the total number of labeled d2 commissural axons. For al., 2004; Reeber et al., 2008). We used CAG-tauGFP and CAG- quantification of spinal cord-derived axons in the brain, the numbers of taumCherry, in which tauGFP and taumCherry sequences from ipsilaterally and contralaterally projecting d1 and d2 axons were counted Neurog1tauGFP and Neurog1taumCherry, respectively, were sub- in the , cerebellum, and around the isthmus at the hind- cloned into the pPB-CAG vector (Yusa et al., 2009), as well as CAG- brain– border. In the hindbrain, only rostrally projecting GPI-AP, which expresses GPI-anchored human placental alkaline axons were counted. Rostrally projecting axons anterior to the cere- phosphatase (AP; Arakawa et al., 2008), as pan neuronal/axonal bellar peduncles of the spinocerebellar tract were counted as axons markers. Ncad and standard control morpholino oligomers (MO) projecting to the isthmus. In each case, axon counts were normalized with 3Ј-fluorescein or lissamine tags were obtained from Gene to the corresponding numbers of all transfected cell bodies in a given Tools. The antisense oligo sequence for the Ncad MO is 5Ј- spinal cord. All data analyses were performed using the unpaired GCGGCGTTCCCGCTATCCGGCACAT-3Ј, which targets the trans- Student’s t test and GraphPad Prism (Version 5.0d), and p Ͻ 0.05 was lational start region of chick Ncad (Shiau and Bronner-Fraser, 2009). considered statistically significant. Sakai et al. • Robo Inhibits Ncad to Form Spinocerebellar Tract J. Neurosci., October 31, 2012 • 32(44):15377–15387 • 15379

Results spinal cord section, ϳ4 cell bodies colabeled with retrogradely d1 and d2 axons project to the chick hindbrain, cerebellum, transferred WGA, whereas ϳ6 cells of 45 labeled Neurong1 and isthmus along MLc and ILc trajectories progenitor-derived neurons were also labeled with WGA. Given After crossing the midline, ascending d1 and d2 dorsal spinal com- that the colabeled cell bodies were located in lamina VI and VII of missural axons contribute to both the VF and LF by projecting along Atoh1CreER;Rosa26-LacZflox (Fig. 2B–D) and Neurog1CreER; MLc and ILc trajectories, respectively, within the spinal cord mar- Rosa26-LacZflox (Fig. 2E–G) mice, respectively, and both laminae ginal zone (Reeber et al., 2008; Avraham et al., 2009) (Fig. 1J). How- contain spinocerebellar neurons in mammals (Brodal, 1998; Xu ever, the brain targets of these neurons have not been explicitly and Grant, 2005), these findings indicate that, just as in chick identified, and whether the positioning of the corresponding axons embryos, d1 and d2 axons contribute to the spinocerebellar tract within particular funiculi in the spinal cord proper influences their in mice. projection to these targets has not been addressed. Toward investi- gating these issues we first performed unilateral in ovo electropora- Overexpression of Ncad results in a reduction of ILc axons tion with specific reporter constructs (Lumpkin et al., 2003; Nakada within the LF, phenocopying the effects of disabling Robo-Slit et al., 2004; Reeber et al., 2008) to identify the brain targets of d1 and signaling d2 neurons/axons in chick embryos. Specifically, we analyzed the Toward determining how the axons of spinal projection neurons long-range projections of d1 and d2 axons in whole-mount prepa- project to their brain targets, we next investigated the molecular rations derived from the electroporated embryos by photoconvert- mechanisms that sort these axons into distinct longitudinal tracts ing the GFP labeling of d1 and d2 axons to more robust colorimetric within the spinal cord proper. As a consequence of disabling Robo- signals (Fig. 1A–H,J). The perdurance of reporter expression made Slit signaling in the chick spinal cord, decussated commissural axons it possible to trace d1 and d2 axons in embryos electroporated at fail to grow away from the ventral midline along ILc trajectories and, E2.5–E3 and harvested as late as E10–E14. The majority of d1 and d2 instead, exclusively extend adjacent to the floor plate as MLc projec- axons projected anteriorly, and these axons were readily visible as far tions in the VF (Reeber et al., 2008) (Fig. 3A,E,H). To explore the rostrally as the isthmus or hindbrain–midbrain border located be- possibility that Robo-mediated inhibition of Ncad (see Introduc- tween the cerebellum and tectum, on both the contralateral and tion) might account for this phenotype, we first investigated the ipsilateral (electroporated) sides of the spinal cord (Fig. 1A-H,J). distribution of Ncad in the chick spinal cord. Ncad is expressed on Importantly, most d1 and d2 axons within the LF projected to the spinal commissural axons, including those extended by genetically cerebellum, forming the inferior portion of the cerebellar peduncles distinct d1 and d2 neurons (data not shown; Fig. 3B,C). These ob- (Fig. 1A–D,G,J), whereas the majority of these axons within the VF projected further rostrally to the isthmus (Fig. 1A–H,J). servations raised the possibility that postcrossing spinal commis- To further characterize the projections of d1 and d2 axons as sural axons hyperfasciculate within the VF in an Ncad-dependent they extend to their brain targets, we quantified the number of manner following the disruption of Robo-Slit signaling (Fig. 3A). A longitudinally projecting axons that reached the hindbrain (Fig. straightforward prediction of this hypothesis is that overexpression 1C,F,I), cerebellum (Fig. 1D,G,I), and the isthmus (Fig. of full-length Ncad (NcadFL) on commissural axons would abro- 1E,H,I), on both the contralateral and ipsilateral sides of the gate Robo-mediated inhibition of Ncad, leading to an increase in electroporated embryos. In all cases, the numbers of labeled ax- MLc axons and a corresponding decrease in ILc axons, effectively ons were normalized to the numbers of transfected neurons (Fig. recapitulating the pathfinding phenotype that results from disrupt- 1I). Of the axons that reached the hindbrain, the small number ing Robo-Slit signaling. (and/or collaterals emanating from these axons), which executed As a first step toward testing this prediction, we asked whether a ventral turn within the hindbrain proper, were excluded from it would be possible to overexpress Ncad on spinal commissural the quantification (Fig. 1C,F). Together, our findings show that axons in vivo via unilateral electroporation of E3–E3.5 chick em- d1 and d2 spinal axons project to at least three distinct brain bryos with a chick NcadFL expression construct and a CAG- regions through the VF and LF (Fig. 1K) and that similar num- tauGFP pan-axonal reporter plasmid. At E4.5, immunostaining bers of these axons project to a given region. with anti-chick Ncad confirmed an increased level of Ncad ex- pression on tauGFP-labeled commissural axons on the trans- Retrograde labeling of spinocerebellar tract: Atoh1 and fected side of the spinal cord (Fig. 3C). We then coexpressed Neurog1 progenitor-derived neurons project to the NcadFL and a pan-axonal marker, CAG-taumCherry, or the d2 cerebellum in mice reporter (Neurog1taumCherry) to assess the consequences of To further confirm that d1 and d2 axons target the cerebellum, overexpressing Ncad on the axons of all transfected neurons or we performed retrograde tracing of spinocerebellar axons in only the d2 subset. It is important to note that electroporation of postnatal Atoh1 (Atoh1CreER;Rosa26-LacZflox) (Soriano, 1999; either the pan-neuronal/axonal or the d2 reporter labels both ILc Machold and Fishell, 2005) and Neurog1 (Neurog1CreER;Rosa26- and MLc projections (Reeber et al., 2008) (Fig. 3D,G). Also, here lacZflox) (Koundakjian et al., 2007) reporter mice following ta- and throughout this study we quantified the various misexpres- moxifen administration at E10.5. Specifically, we injected Alexa sion phenotypes by selectively counting the numbers of d2 axons. Fluor 555-conjugated WGA into the cerebellum of P5 Three days after coelectroporating E3 chick embryos with Atoh1CreER;Rosa26-LacZflox and Neurog1CreER;Rosa26-lacZflox NcadFL and pan-axonal or d2 reporter constructs we observed a mice. Subsequently, X-gal staining and/or ␤-gal expression in significant reduction in the numbers of ILc axons and an abnor- spinal cord sections obtained from these animals at P6 was used mally thick MLc axon bundle in open-book preparations derived to visualize fate-mapped Atoh1 progenitor-derived d1 and Neu- from these embryos (Fig. 3F,I). Consistent with a potential role rog1 progenitor-derived neurons (Fig. 2A,B,D,E,G). These anal- for Robo inhibition of Ncad adhesion in directing a subset of yses revealed that a subset of ␤-gal-expressing Atoh1- and postcrossing commissural axons along ILc trajectories and into Neurog1-derived cell bodies was colabeled with the retrogradely the LF, these defects resemble those observed in embryos electro- transported WGA tracer in the spinal cord (Fig. 2B–G). Among 9 porated with a cytoplasmic truncation (dominant-negative ␤-gal-positive d1 neurons contained within one side of a given form) of Robo, CAG-Robo1⌬C-GFP (Fig. 3E,H). 15380 • J. Neurosci., October 31, 2012 • 32(44):15377–15387 Sakai et al. • Robo Inhibits Ncad to Form Spinocerebellar Tract

We quantified the NcadFL phenotype by analyzing the pathfinding defects displayed by d2 commissural axons in further detail. First, we confirmed that similar numbers of d2 commissural axons, normalized to the numbers of labeled cell bodies, crossed the floor plate (Fig. 3J,K, “crossing axons”) in embryos electroporated with the reporter plasmids alone (control) and coelectropo- rated with the reporter and c␤actin-NcadFL plasmids (NcadFL; Fig. 3J). This confirmed that similar transfection efficiencies were achieved in control and experimental electroporations and revealed that d2 ax- ons successfully crossed the floor plate fol- lowing misexpression of NcadFL. We next quantified the postcrossing com- missural axon phenotype resulting from the misexpression of NcadFL by deter- mining the ratio of d2 ILc axons to de- cussated d2 axons. Consistent with our qualitative observations, the number of d2 ILc axons is significantly reduced in embryos electroporated with c␤actin- NcadFL, compared with control condi- tions (Fig. 3K). Collectively, our findings suggest that the hyperfasciculation of postcrossing MLc axons resulting from the misexpression of NcadFL prevents de- cussated commissural axons from pro- jecting away from the floor plate along ILc trajectories and joining the LF.

Loss of Ncad function alone does not perturb the pathfinding of postcrossing commissural axons Our findings raised the possibility that Ncad is normally required for the fascicu-

4

collaterals are shown. d1 (D) and d2 (G) commissural axons form peduncles (black arrows) directed toward the cerebel- lum. White arrows indicate d1 (E) and d2 (H) commissural axons projecting to the isthmus. Midbrain (Mb) tectum is re- moved for visual clarity. I, Quantification of d1 (n ϭ 4) and d2 (nϭ4)axonsprojectinginthehindbrain,thecerebellum,and isthmus on the contralateral and ipsilateral (electroporated) sides of the E10 chick brain. J, d2 axons in the ventral (top down) view of whole-mount E10 chick spinal cord and brain. Spinal cord is shown in open-book view. White line indicates the transfected area on the electroporated side. White arrows point to the two hemispheres of the cerebellum. In the spinal cord, the left boxed region locates at the thoracic level in the contralateral side, and the right box covers the cervical level. Black arrows indicate contralaterally projecting ILc axons. In thehindbrain,redarrowspointtotheaxonswithintheLFthat target the cerebellum, and green arrows point to the axons projecting to the isthmus from the VF. K, Schematic of spinal commissural axon trajectories in open-book view of the spinal Figure 1. d1 and d2 axons project to the hindbrain, cerebellum, and isthmus in chick embryos. A, B, Lateral view of E10.5 (A) cord and brain. Red axons follow ILc trajectories contributing and E11.5 (B) chick brain shows d1 (A) and d2 (B) commissural axons projecting to the hindbrain (Hb), the cerebellum (Cb; black to the LF and projecting to the cerebellum. On the other hand, arrows), and the isthmus (white arrows). Arrowheads in the hindbrain indicate the axons and/or collaterals, which make orthog- thegreenaxonsadoptanMLctrajectory,jointheVF,andproj- onalturnsandprojectventrally.C–H,LateralviewofE11chickbrainshowsd1(C–E)andd2(F–H)commissuralaxonsinthebrain. ect to, and most likely beyond, the isthmus. fp, floor plate; rp, In the hindbrain (C, F), rostrally projecting and ventrally terminating (arrowheads) d1 (C) and d2 (F) commissural axons/ roof plate; P, posterior; A, anterior; D, dorsal; V, ventral. Sakai et al. • Robo Inhibits Ncad to Form Spinocerebellar Tract J. Neurosci., October 31, 2012 • 32(44):15377–15387 • 15381

disrupt the pathfinding of postcrossing commissural axons (data not shown). Together, these findings indicate that the loss of Ncad function alone does not interfere with the directed growth of postcrossing spinal commissural axons into the VF or LF.

Simultaneous perturbation of Ncad and Robo function restores ILc axons and the contralateral lateral funiculus Our findings to this point are consistent with a mechanism in which Robo normally inhibits Ncad function on postcrossing commissural axons and that this selectively facilitates the formation of the ILc axon- containing LF. To provide further support for this model, we co-electroporated E3– E3.5 chick embryo spinal cords with both CAG-Robo1⌬C-GFP and NcadMO and as- sessed the consequences of simultaneously perturbing Ncad and Robo function on the formation of postcrossing MLc and ILc ax- ons. As revealed by the expression of GFP in open-book preparations derived from these embryos, interfering with both Robo-Slit signaling and Ncad partially rescued the dominant-negative Robo phenotype (Fig. 5A,B). d2 ILc axons were also rescued to a similar extent in the absence of both Robo and Ncad function (Fig. 5C,D). Quantifica- tion of these phenotypes revealed a signifi- cant increase in the number of d2 ILc axons within spinal cords transfected with both CAG-Robo1⌬C-GFP and Ncad MO, as compared with those electroporated with the dominant-negative Robo1 construct and control MO (Fig. 5I). Utilizing a complementary strategy to Figure 2. Retrograde tracing of spinocerebellar tract labels Atoh1 progenitor-derived d1 and Neurog1-lineage neurons in eliminate Ncad function we misexpressed mouse spinal cord. A–D, P6 spinal cord of Atoh1CreER;Rosa26-LacZflox mouse showing fate-mapped d1 cell bodies stained with ␤ a dominant-negative form of Ncad, CAG- X-gal (A). Arrows identify the d1 cell body colabeled with -gal and a retrograde tracer, Alexa Fluor 555 conjugate of WGA, which ⌬ was injected into the cerebellum at P5 (B–D). D, High-magnification views and merged image of boxed regions in B and C. E–G, Ncad E. This plasmid encodes an extra- P6 spinal cord of Neurog1CreER;Rosa26-lacZflox mouse showing a Neurog1-derived cell body colabeled with WGA (see arrows). G, cellular truncation of Ncad, which is High-magnification views and merged image of boxed areas in E and F. incapable of interacting with NcadFL, but can bind downstream signaling molecules such as ␤-catenin via its cytoplasmic do- lation of postcrossing MLc axons and, subsequently, VF forma- main and, thus, represents a dominant-negative form of Ncad tion. We tested this hypothesis by carrying out a variety of loss- (Bozdagi et al., 2004). Consistent with the phenotype of chick of-function manipulations. First, we used an MO to knock down embryos transfected with both CAG-Robo1⌬C-GFP and Ncad protein expression and assessed the consequences of this NcadMO, co-electroporation of dominant-negative Robo1 and manipulation on commissural axon pathfinding. Consistent with Ncad constructs partially rescued ILc projections (Fig. 5E-H). a previous report (Shiau and Bronner-Fraser, 2009), transfection Together with the NcadFL overexpression phenotypes, these of Ncad MO significantly reduced Ncad expression in the chick findings suggest that Robo-mediated inhibition of Ncad nor- spinal cord (Fig. 4A,B). If Ncad-mediated fasciculation is re- mally facilitates the formation of ILc axons and, consequently, quired for the formation of MLc axon tracts, the loss of Ncad the LF, by preventing the hyperfasciculation of postcrossing ax- might be expected to result in a decrease in the number of MLc ons within the VF. axons and a concomitant increase in the number of ILc axons. However, after transfecting E3–E3.5 chick spinal cord with the Knockdown of Ncad via in utero electroporation in Ncad Ncad MO and pan-axonal or d2 reporters we observed no obvi- conditional mutant mice rescues the Robo⌬C perturbation ous alterations in the numbers of ILc and MLc axons (Fig. 4C–F). To provide further support for Robo-mediated inhibition of Ncad In separate experiments, we found that electroporation of sev- function regulating contralateral commissural projections, and to eral different dominant-negative Ncad expression constructs determine whether this mechanism is conserved in mammals, we capable of perturbing Ncad function (see below) also failed to used in utero electroporation to ask whether interfering with Ncad 15382 • J. Neurosci., October 31, 2012 • 32(44):15377–15387 Sakai et al. • Robo Inhibits Ncad to Form Spinocerebellar Tract

Figure3. MisexpressionofNcadFLonspinalcommissuralaxonsleadstoareductioninthenumberofILcaxons.A,Schematicofcommissuralaxontrajectoriesinopen-bookviewofthespinalcordunilaterally electroporatedwithreporterconstructs.Axonsfromredandgreenneuronscrossthefloorplate(fp)andelaboratecontralateralprojections.Inourmodel,RoboinhibitionofNcaddirectsredaxontoelaboratean ILc trajectory and join the LF. Perturbing Robo function forces red axon to grow exclusively alongside the fp within the VF as a MLc projection. a, anterior; p, posterior; rp, roof plate. B, Ncad expression on d2 commissural axons in E4.5 chick spinal cord. Arrows identify precrossing segments of commissural axons. Arrowheads point to the ventral commissure. C, Overexpression of NcadFL in E5 chick spinal cord followingelectroporationofc␤actin-NcadFLandCAG-tauGFPconstructsatE3.WhitearrowsindicatemisexpressedNcadonprecrossing,andyellowarrowspointtopostcrossingsegmentsofGFP ϩcommissural axons.Arrowheadsidentifytheventralcommissure.D–I,Decussatedaxonslabeledwiththepan-axonalmarkerCAG-taumCherry(D–F)ord2markerNeurog1taumCherry(G–I)onthecontralateralsideofthe electroporatedembryoareshowninopen-bookview.Comparedwithcontrolembryos(D,G),misexpressionofNcadFLresultsinfewerILcaxonsandlargerMLcbundles(F,I),phenocopyingtheeffectofdisabling Robofunction(E,H)inE6chickspinalcord.Scalebars:D(forD–F),G(forG–I),100␮m.J,K,Quantificationofd2ILcaxonphenotypeinwild-typeandNcadFL-misexpressingembryos.Thedifferencebetween normalized counts of d2 commissural axons in control (n ϭ 5) and experiment (n ϭ 5) conditions is not statistically significant, (J) whereas normalized counts of d2 ILc axons misexpressing NcadFL are significantly reduced as compared with controls (K,*pϽ0.05). All values show mean ϮSEM. function rescues the dominant-negative Robo phenotype in the Ncadfl/fl embryos with CAG-Robo1⌬C-GFP and CAG-IRES-GFP pre- mouse spinal cord. Since Ncad homozygous-null mutant mice die vented the formation of ILc axons (Fig. 6B). On the other hand, co- by E10 (Radice et al., 1997), we electroporated floxed-Ncad electroporation of CAG-Robo1⌬C-GFP and CAG-nlsCre-IRES-GFP, (Ncadflox) mice (Kostetskii et al., 2005) with a Cre expression con- which encodes a Cre sequence with a nuclear localization signal, in struct to selectively eliminate Ncad in the spinal cord of E11.5 ho- Ncadfl/fl embryos, rescued the dominant-negative Robo phenotype mozygous embryos (Ncadfl/fl). Just as we observed in the chick spinal and resulted in the formation of wild-type-like ILc axons (Fig. cord, commissural axons transfected with the general reporter con- 6C,D). These findings provide further support for Robo-mediated struct, CAG-tauGFP, and a control plasmid, CAG-IRES-GFP, which inhibition of Ncad regulating the positioning of longitudinal spinal lacks a Cre sequence, elaborated both ILc and MLc trajectories in the axon tracts and indicate that this mechanism is conserved in Ncadfl/fl embryos (Fig. 6A), whereas, unilateral co-electroporation of mammals. Sakai et al. • Robo Inhibits Ncad to Form Spinocerebellar Tract J. Neurosci., October 31, 2012 • 32(44):15377–15387 • 15383

Figure 4. Knockdown of Ncad alone does not perturb the postcrossing trajectories of spinal commissural axons. A, B, Ncad expression in the E5 chick spinal cord is reduced (see arrows) after transfection of 3Ј-lissamine-tagged Ncad MO (B), as compared with embryos transfected with control MO (A). Scale bar, 50 ␮m. C–F, Ncad knockdown (D, F) results in normal ILc and MLc projections, similar to controls (C, E), as shown by axons labeled with the pan-axonal (C, D) or d2 axon (E, F) reporters in E6 chick spinal cord. Scale bar, 100 ␮m.

Figure 5. Knockdown of Ncad or mis-expression of a dominant-negative form of Ncad partially rescues the Robo⌬C phenotype. A–D, Cotransfection of E3–E3.5 chick spinal cord with Ncad MO and CAG-Robo1⌬C-GFP (B, D) restores ILc projections compared with controls (A, C) at E6, as shown by GFP-tagged Robo1⌬C-expressing axons (A, B) and d2 axons (C, D). E–H, Electroporation of E3–E3.5chickspinalcordwithCAG-Ncad⌬EandCAG-Robo1⌬C-GFP(F,H)alsoresultsintherescueofRobo⌬CphenotypeinE6embryos,comparedwithcontrolconditions(E,G).Axonsarelabeled with Robo⌬C-GFP (E, F) or d2 reporter (G, H) constructs. Scale bar, 100 ␮m. I, The normalized counts of d2 ILc axons are significantly increased (mean Ϯ SEM, *p Ͻ 0.05) in chick spinal cords transfected with Ncad MO and CAG-Robo1⌬C-GFP (n ϭ 3) compared with the control (n ϭ 3).

Figure6. NcadknockdownrestoresILctrajectoryinNcadfl/fl mice.A–C,E13.5open-bookpreparationsfromNcadfl/fl micespinalcordelectroporatedinuteroatE11.5showILcandMLcprojections of spinal commissural axons expressing reporter constructs only (A), a reduction in ILc axons following misexpression of Robo⌬C(B), and the rescue of ILc axons expressing Robo⌬C and Cre (C). D, SignificantincreaseinILcaxons(meanϮSEM,**pϽ0.005)withinNcadfl/fl spinalcordelectroporatedwithCAG-Robo1⌬C-GFPandCAG-nlsCre-IRES-GFP(nϭ3),ascomparedwithNcadfl/fl spinal cord transfected with CAG-Robo1⌬C-GFP and control plasmids (n ϭ 3).

Robo function or inhibition of Ncad-mediated adhesion is geting. Since disabling Robo-Slit signaling results in the loss of required for the targeting of spinocerebellar axons in the ILc axons, and the commissural axon-containing portion of the brain LF (Fig. 3E,H), we exploited this manipulation to selectively ask Given that spinal commissural axons project to multiple brain whether mis-sorted decussated spinal commissural axons project regions, including the cerebellum, we asked whether the posi- to their appropriate brain targets. For these analyses, we used a tioning of these axons in distinct longitudinal tracts within the general axon reporter construct CAG-GPI-AP (Arakawa et al., spinal cord marginal zone is required for their proper brain tar- 2008), which expresses GPI-anchored human placental AP, to 15384 • J. Neurosci., October 31, 2012 • 32(44):15377–15387 Sakai et al. • Robo Inhibits Ncad to Form Spinocerebellar Tract

Figure7. SpinalcommissuralaxonsexpressingRobo1⌬CorNcadFLfailtoprojecttothecerebellumandmis-projecttotheisthmus,andperturbationofNcadrescuesthisphenotype.A–D,Spinal commissural axons on the contralateral side of E13 whole-mount chick brain following transfection of E3–E3.5 spinal cord with respective constructs and MO. As compared with wild-type embryos (A), the contralateral spinocerebellar tract (black arrowhead with asterisk) is reduced in embryos electroporated with CAG-Robo1⌬C-GFP and control MO, whereas the projection of axons to the isthmus (white arrowheads) is increased (B). Simultaneous perturbation of Robo and Ncad function restores the projection of spinal commissural axons to the cerebellum (Cb; indicated by two asterisks in C). A subset of spinal commissural axons misexpressing NcadFL also fails to project to the cerebellum (D). E–H, In E10 chick brain, there is also a reduction in the projection of Robo1⌬C or NcadFL-expressing d2 commissural axons to the cerebellum (asterisk in F,H) compared with the control (E). Cotransfection of CAG-Robo1⌬C-GFP and Ncad MO in the E3 chick spinal cord restores the d2 spinocerebellar tract (two asterisks in G). I, Normalized d2 axon counts in the hindbrain, the cerebellum, and the isthmus indicate a significant reduction (**p Ͻ 0.01) in the spinocerebellar tract and a significant increase in projection to the isthmus (*p Ͻ 0.05) formed by d2 commissural axons following transfection with CAG-Robo1⌬C-GFP and control MO (Robo1⌬C; n ϭ 3), compared with the control (wild-type, WT; n ϭ 3). Normalized axon counts of d2 commissural axons misexpressing NcadFL (NcadFL; n ϭ 3) are also significantly reduced in the cerebellum, comparedwiththecontrol(*pϽ0.05).Thereisnostatisticallysignificantdifferencebetweennormalizedd2commissuralaxoncountsinWTembryosandthosetransfectedwithRobo1⌬C-GFPand Ncad MO (Robo1⌬C ϩ NcadMO; n ϭ 3), indicating the rescue of projections to the cerebellum and isthmus. All values show mean Ϯ SEM. Hb, hindbrain; Mb, midbrain. achieve robust colorimetric labeling of axons in a subset of em- projecting to the cerebellum was significantly reduced in em- bryos, and the d2 reporter construct for quantifying the targeting bryos transfected with the dominant-negative Robo1 construct, of labeled spinal commissural axons in a separate set of embryos. (Fig. 7B,F,I), compared with control embryos (Fig. 7A,E,I). On The brains of the embryos electroporated with the pan-axonal the other hand, the number of commissural axons projecting to and d2-specific reporters were analyzed at E13 and E10, respec- the isthmus was significantly increased in the embryos trans- tively, after co-electroporation of the E3–E3.5 chick spinal cord fected with CAG-Robo1⌬C-GFP (Fig. 7B,F,I). Further consistent with either control CAG-tauGFP or CAG-Robo1⌬C-GFP con- with Robo-mediated inhibition of Ncad regulating the targeting structs. The numbers of the transfected d2 cell bodies did not vary of spinal projection neurons, simultaneously perturbing Robo significantly between experiments: 3419 Ϯ 266.3 for control, and Ncad function rescued the contralateral spinocerebellar tract 3214 Ϯ 74.27 for CAG-Robo1⌬C-GFP, 2509 Ϯ 191.5 for CAG- (Fig. 7C,G,I). Importantly, a subset of spinal commissural axons Robo1⌬C-GFP and NcadMO, and 3928 Ϯ 52.78 for NcadFL (n ϭ that overexpress NcadFL also fail to project to the cerebellum 3 for each experiment). Notably, the normalized counts of ros- (Fig. 7D,H,I). trally projecting axons in the hindbrain were similar in embryos Together, these results suggest that Robo-mediated inhibition electroporated with the control or CAG-Robo1⌬C-GFP con- of Ncad adhesion is required for the proper positioning of post- structs, indicating that spinal axons appropriately reached the crossing spinal commissural axons within the LF of the spinal hindbrain in the presence or absence of Robo signaling function cord marginal zone and, in turn, for their targeting to the cere- (Fig. 7A–C,E–G,I). However, the number of commissural axons bellum (Fig. 8). Sakai et al. • Robo Inhibits Ncad to Form Spinocerebellar Tract J. Neurosci., October 31, 2012 • 32(44):15377–15387 • 15385

and MLc axons into the LF and VF, re- spectively (Reeber et al., 2008; Avraham et al., 2009). We show here that, consistent with the known anatomy and physiology of spinal projection neurons, d1 and d2 ax- ons project to multiple brain regions, and that a significant subset of both populations targets the cerebellum. This latter observa- tion is consistent with the results of our ret- rograde labeling experiments performed in Atoh1 and Neurog1 reporter mice, as well as with the findings of an independent dye-tracing study suggesting that Atoh1 progenitor-derived d1 neurons/axons con- tribute to the spinocerebellar tract (Ber- mingham et al., 2001). In addition, we show for the first time that d1 and d2 axons also contribute to other major ascending tracts, which project to the brain. For example, some d1 and d2 axons and/or their collater- als in the hindbrain execute ventrally di- rected orthogonal turns, identifying them as likely components of spino-olivary and spinoreticular tracts (Kerr, 1975; Brodal, Figure 8. Sorting of postcrossing commissural axons via Robo-mediated inhibition of Ncad is required for spinocerebellar tract 1998). The remaining sets of labeled axons formation.Schematicofspinalcommissuralaxontrajectoriesinopen-bookviewofthespinalcordandbrain.Inwild-typeembryos, appear to be destined for the cerebellum Robo inhibition of Ncad directs red axons along an ILc trajectory, so that they can join the LF and project to the cerebellum (Cb). On or more rostrally located targets. How- the other hand, the green axons join the VF by adopting an MLc trajectory and project to, and most likely beyond, the isthmus. As ever, we were unable to follow these par- a consequence of perturbing Robo function (Robo loss of function), red axons inappropriately elaborate MLc projections that grow ticular axons beyond the isthmus, even exclusively alongside the floor plate (fp) within the VF and which, in turn, mis-project to the isthmus. rp, roof plate; Hb, hindbrain; using a pan-axonal reporter that provides Mb, midbrain. more robust labeling of these projections. This is most likely due to the fact that after reaching the isthmus ascending spinal ax- Discussion ons, which contribute to the anterolateral system, are no longer Here, we describe, for the first time, a molecular mechanism that visible as they invade deep interior regions of the midbrain (Kerr, controls the directed growth of spinal axons to brain targets. Accord- 1975; Bjo¨rkeland and Boivie, 1984; Brodal, 1998). ing to our model, the axons of spinal commissural projection neu- rons are presorted into distinct longitudinal funiculi within the A critical role for axon sorting regulated by Robo-mediated marginal zone, via Robo-mediated inhibition of Ncad, and this fa- inhibition of Ncad-dependent fasciculation in spinocerebellar cilitates their projection to appropriate brain targets (Fig. 8). tract formation Although it is well established that ascending longitudinal axons d1 and d2 projection neurons target multiple brain regions are positioned within distinct funiculi in the spinal cord marginal Anatomical and physiological studies in vertebrates have corre- zone, whether this organizational pattern is required for the long- lated the positions of funiculi within the spinal cord marginal range guidance and targeting of these axons has not been ad- zone with the particular brain regions that their component ax- dressed. Our findings suggest that ascending projection neuron ons ultimately target (Brodal, 1998; Willis, 2007; Sakai and Kapri- axons must be presorted within the spinal cord proper to project elian, 2012). In particular, the spinocerebellar tract is the only to their appropriate brain targets. Similarly, olfactory sensory major ascending projection in the dorsal LF that targets the brain axons are spatially arranged according to their target sites in the (Brodal, 1998; Xu and Grant, 2005; Willis, 2007). The ventral LF olfactory bulb and this topographic organization is, in part, reg- harbors additional components of the spinocerebellar tract and a ulated by the molecular identities of individual olfactory sensory portion of the anterolateral system, and the VF contains the re- neurons, the expression of distinct olfactory receptors, and, ulti- maining portion of the anterolateral system and the spino-olivary mately, by axon-associated expression gradients of canonical tract (Kerr, 1975; Giesler Jr. et al., 1981; Richmond et al., 1982; guidance cues such as Semaphorins and Neuropilins (Sakano, Brodal, 1998; Xu and Grant, 2005; Willis, 2007). The spatial or- 2010). Notably, presorting of olfactory sensory axons, as opposed ganization of longitudinal tracts within the marginal zone re- to target-derived guidance cues, regulates the targeting of these vealed by these studies suggests that spinal projection neurons axons within the olfactory bulb (Imai et al., 2009). (including d1 and d2) whose axons contribute to the LF and VF, Analogous to the mechanisms underlying axon targeting in connect to multiple brain regions, and that spinal ascending ax- the olfactory system, we propose that Robo inhibition of Ncad ons, which join the dorsal LF, project to the cerebellum. How- function sorts decussated spinal commissural axons into the LF ever, the brain targets of specific subtypes of projection neurons within the spinal cord marginal zone, and that this is required for have yet to be explicitly identified. their correct projection to the cerebellum. Our findings suggest In the chick spinal cord, it has previously been shown that that the high levels of Robo likely expressed on ILc axons inhibit both d1 and d2 commissural neurons extend postcrossing ILc Ncad-dependent fasciculation, allowing these axons to break 15386 • J. Neurosci., October 31, 2012 • 32(44):15377–15387 Sakai et al. • Robo Inhibits Ncad to Form Spinocerebellar Tract away from the VF and join the LF. In support of this model, References simultaneous perturbation of Robo and Ncad function in the Arakawa T, Iwashita M, Matsuzaki F, Suzuki T, Yamamoto T (2008) Paths, spinal cord partially rescues the formation of not only the con- elongation, and projections of ascending chick embryonic spinal com- tralateral LF, but also the spinocerebellar tract. Although it is missural neurons after crossing the floor plate. Brain Res 1223:25–33. certainly possible that Robo and Ncad regulate the formation of CrossRef Medline Avraham O, Hadas Y, Vald L, Zisman S, Schejter A, Visel A, Klar A (2009) ILc axons and the LF through distinct parallel signaling pathways, Transcriptional control of axonal guidance and sorting in dorsal in- we favor our model since disabling Robo function has been dem- terneurons by the Lim-HD proteins Lhx9 and Lhx1. Neural Dev 4:21. onstrated to eliminate virtually all ILc axons (Long et al., 2004; CrossRef Medline Reeber et al., 2008; Jaworski et al., 2010) (Figs. 3, 5, 6). Moreover, Bermingham NA, Hassan BA, Wang VY, Fernandez M, Banfi S, Bellen HJ, elegant biochemical studies have shown that Slit-activation of Fritzsch B, Zoghbi HY (2001) Proprioceptor pathway development is Robo inhibits Ncad-mediated adhesion in vitro (Rhee et al., 2002, dependent on MATH1. Neuron 30:411–422. CrossRef Medline 2007), and Robo-Slit signaling facilitates retinal ganglion cell de- Bjo¨rkeland M, Boivie J (1984) The termination of spinomesencephalic fi- bers in cat. Anat Embryol 170:265–277. CrossRef Medline velopment by attenuating Ncad activity in zebrafish (Wong et al., Bovolenta P, Dodd J (1990) Guidance of commissural growth cones at the 2012). floor plate in embryonic rat spinal cord. Development 109:435–447. Consistent with the Robo-Ncad model we propose, VF forma- Medline tion, likely driven by the fasciculation of MLc axons, might also Bozdagi O, Valcin M, Poskanzer K, Tanaka H, Benson DL (2004) Tempo- be dependent on Ncad function. Accordingly, the selective dis- rally distinct demands for classic cadherins in synapse formation and ruption of Ncad could result in a defasciculation of MLc axon maturation. Mol Cell Neurosci 27:509–521. CrossRef Medline bundles, effectively reducing the thickness of the VF. However, Brodal P (1998) The central nervous system: structure and function. New York: Oxford UP. we show that introducing Ncad MOs into, or misexpressing a Bultje RS, Castaneda-Castellanos DR, Jan LY, Jan YN, Kriegstein AR, Shi SH dominant-negative Ncad construct in, the embryonic chick spi- (2009) Mammalian Par3 regulates progenitor cell asymmetric division nal cord does not significantly perturb the pathfinding of post- via notch signaling in the developing neocortex. Neuron 63:189–202. crossing commissural axons. Whereas the lack of an axon-sorting CrossRef Medline phenotype in these experiments could reflect an incomplete or Garel S, Rubenstein JL (2004) Intermediate targets in formation of topo- insufficient disruption of Ncad function, it is also possible that graphic projections: inputs from the thalamocortical system. Trends Neurosci 27:533–539. CrossRef Medline Ncad is specifically expressed on the ILc subset of postcrossing Giesler GJ Jr, Spiel HR, Willis WD (1981) Organization of spinothalamic commissural axons and not required for the directed growth of tract axons within the rat spinal cord. J Comp Neurol 195:243–252. MLc axons. However, it would be difficult to visualize such an CrossRef Medline axon subtype-specific expression pattern since ILc axons must Gilthorpe JD, Papantoniou EK, Che´dotal A, Lumsden A, Wingate RJ (2002) project through the VF to reach the LF. Alternatively, Ncad may The migration of cerebellar rhombic lip derivatives. Development 129: have a permissive role in longitudinal axon guidance and tract 4719–4728. Medline formation. In this case, the consequences of disrupting Ncad may Hammond R, Vivancos V, Naeem A, Chilton J, Mambetisaeva E, Andrews W, Sundaresan V, Guthrie S (2005) Slit-mediated repulsion is a key regula- only be discernible in the absence of Robo function. Since axon tor of motor axon pathfinding in the hindbrain. Development 132:4483– fasciculation and outgrowth generally require the concerted ac- 4495. CrossRef Medline tions of a variety of cell adhesion molecules (Van Vactor, 1998; Imai T, Yamazaki T, Kobayakawa R, Kobayakawa K, Abe T, Suzuki M, Sakano Raper and Mason, 2010), the selective elimination of Ncad func- H (2009) Pre-target axon sorting establishes the neural map topogra- tion might not be expected to result in a clear phenotype. Poten- phy. Science 325:585–590. CrossRef Medline tially relevant to either scenario, it is interesting to note that the Imondi R, Kaprielian Z (2001) Commissural axon pathfinding on the con- spinal cord of mouse embryos lacking the atypical cadherin, tralateral side of the floor plate: a role for B-class ephrins in specifying the dorsoventral position of longitudinally projecting commissural axons. Celsr3, which is the mammalian homolog of Drosophila Fla- Development 128:4859–4871. Medline mingo, exhibit an abnormally thick LF and a thinner VF. More- Imondi R, Wideman C, Kaprielian Z (2000) Complementary expression of over, the spinocerebellar tract, which travels within the dorsal LF, transmembrane ephrins and their receptors in the mouse spinal cord: a in the hindbrain is also enlarged in Celsr3 mutants (Tissir et al., possible role in constraining the orientation of longitudinally projecting 2005). These phenotypes suggest that Celsr3 has a key role in axons. Development 127:1397–1410. Medline sorting spinal projection neuron axons along MLc trajectories Iwai Y, Usui T, Hirano S, Steward R, Takeichi M, Uemura T (1997) Axon within the VF and their proper projection to brain targets, most patterning requires DN-cadherin, a novel neuronal adhesion receptor, in the Drosophila embryonic CNS. Neuron 19:77–89. CrossRef Medline likely to and beyond the isthmus. Jaworski A, Long H, Tessier-Lavigne M (2010) Collaborative and special- Notably, Ncad expression levels on ascending axons within ized functions of Robo1 and Robo2 in spinal commissural axon guidance. ventrolateral funiculi in the chick spinal cord begin to decrease at J Neurosci 30:9445–9453. Medline E6 and are extinguished at E7 when spinocerebellar tracts reach Jevince AR, Kadison SR, Pittman AJ, Chien CB, Kaprielian Z (2006) Distri- the brain (Redies et al., 1992). As opposed to a direct role for bution of EphB receptors and ephrin-B1 in the developing vertebrate Robo-Ncad interactions in the targeting of longitudinal tracts to spinal cord. J Comp Neurol 497:734–750. CrossRef Medline the brain, this observation supports our contention that Robo- Kadison SR, Kaprielian Z (2004) Diversity of contralateral commissural projections in the embryonic rodent spinal cord. J Comp Neurol 472:411– mediated inhibition of Ncad in the spinal cord proper regulates 422. CrossRef Medline the formation of the LF, which in turn facilitates spinocerebellar Kerr FW (1975) The ventral and other ascending sys- tract formation. Although we favor the view that sorting projec- tems of the ventral funiculus of the spinal cord. J Comp Neurol 159:335– tion of these axons to the brain, target-derived cues may also have 356. CrossRef Medline a role in this process. It is interesting to note in this regard that Kostetskii I, Li J, Xiong Y, Zhou R, Ferrari VA, Patel VV, Molkentin JD, Slits are present in the chick hindbrain and cerebellum when Radice GL (2005) Induced deletion of the N-cadherin gene in the heart leads to dissolution of the intercalated disc structure. Circ Res 96:346– spinocerebellar tracts enter the brain (Gilthorpe et al., 2002). 354. CrossRef Medline Whether or not these brain-derived Slits are also required for Koundakjian EJ, Appler JL, Goodrich LV (2007) Auditory neurons make spinocerebellar tract formation remains an untested and open stereotyped wiring decisions before maturation of their targets. J Neurosci question. 27:14078–14088. CrossRef Medline Sakai et al. • Robo Inhibits Ncad to Form Spinocerebellar Tract J. Neurosci., October 31, 2012 • 32(44):15377–15387 • 15387

Krull CE (2004) A primer on using in ovo electroporation to analyze gene Activation of the repulsive receptor Roundabout inhibits N-cadherin- function. Dev Dyn 229:433–439. CrossRef Medline mediated cell adhesion. Nat Cell Biol 4:798–805. CrossRef Medline Long H, Sabatier C, Ma L, Plump A, Yuan W, Ornitz DM, Tamada A, Mu- Rhee J, Buchan T, Zukerberg L, Lilien J, Balsamo J (2007) Cables links rakami F, Goodman CS, Tessier-Lavigne M (2004) Conserved roles for Robo-bound Abl kinase to N-cadherin-bound beta-catenin to mediate slit and robo proteins in midline commissural axon guidance. Neuron Slit-induced modulation of adhesion and transcription. Nat Cell Biol 42:213–223. CrossRef Medline 9:883–892. CrossRef Medline Lumpkin EA, Collisson T, Parab P, Omer-Abdalla A, Haeberle H, Chen P, Richmond FJ, Courville J, Saint-Cyr JA (1982) Spino-olivary projections Doetzlhofer A, White P, Groves A, Segil N, Johnson JE (2003) Math1- from the upper cervical spinal cord: an experimental study using autora- driven GFP expression in the developing nervous system of transgenic diography and horseradish peroxidase. Exp Brain Res 47:239–251. mice. Gene Expr Patterns 3:389–395. CrossRef Medline Medline Luo L, Flanagan JG (2007) Development of continuous and discrete neural Sakai N, Kaprielian Z (2012) Guidance of longitudinally projecting axons in maps. Neuron 56:284–300. CrossRef Medline the developing central nervous system. Front Mol Neurosci 5:59. Medline Machold R, Fishell G (2005) Math1 is expressed in temporally discrete pools Sakano H (2010) Neural map formation in the mouse olfactory system. of cerebellar rhombic-lip neural progenitors. Neuron 48:17–24. CrossRef Neuron 67:530–542. CrossRef Medline Medline Shiau CE, Bronner-Fraser M (2009) N-cadherin acts in concert with Slit1- Nakada Y, Parab P, Simmons A, Omer-Abdalla A, Johnson JE (2004) Sepa- Robo2 signaling in regulating aggregation of placode-derived cranial sen- rable enhancer sequences regulate the expression of the neural bHLH sory neurons. Development 136:4155–4164. CrossRef Medline transcription factor neurogenin 1. Dev Biol 271:479–487. CrossRef Soriano P (1999) Generalized lacZ expression with the ROSA26 Cre re- Medline porter strain. Nat Genet 21:70–71. CrossRef Medline Niwa H, Yamamura K, Miyazaki J (1991) Efficient selection for high- Tissir F, Bar I, Jossin Y, De Backer O, Goffinet AM (2005) Protocadherin expression transfectants with a novel eukaryotic vector. Gene 108:193– Celsr3 is crucial in axonal tract development. Nat Neurosci 8:451–457. 199. CrossRef Medline Radice GL, Rayburn H, Matsunami H, Knudsen KA, Takeichi M, Hynes RO Medline (1997) Developmental defects in mouse embryos lacking N-cadherin. Van Vactor D (1998) Adhesion and signaling in axonal fasciculation. Curr Dev Biol 181:64–78. CrossRef Medline Opin Neurobiol 8:80–86. CrossRef Medline Raper J, Mason C (2010) Cellular strategies of axonal pathfinding. Cold Willis WD Jr (2007) The , with emphasis on struc- Spring Harb Perspect Biol 2:a001933. CrossRef Medline tures important for . Brain Res Rev 55:297–313. CrossRef Medline Redies C, Inuzuka H, Takeichi M (1992) Restricted expression of N- and Wong GK, Baudet ML, Norden C, Leung L, Harris WA (2012) Slit1b-Robo3 R-cadherin on neurites of the developing chicken CNS. J Neurosci 12: signaling and N-cadherin regulate apical process retraction in developing 3525–3534. Medline retinal ganglion cells. J Neurosci 32:223–228. CrossRef Medline Reeber SL, Sakai N, Nakada Y, Dumas J, Dobrenis K, Johnson JE, Kaprielian Xu Q, Grant G (2005) Course of spinocerebellar axons in the ventral and Z (2008) Manipulating Robo expression in vivo perturbs commissural lateral funiculi of the spinal cord with projections to the posterior cere- axon pathfinding in the chick spinal cord. J Neurosci 28:8698–8708. bellar termination area: an experimental anatomical study in the cat, CrossRef Medline using a retrograde tracing technique. Exp Brain Res 162:250–256. Reeber SL, Gebre SA, Sillitoe RV (2011) Fluorescence mapping of afferent CrossRef Medline topography in three dimensions. Brain Struct Funct 216:159–169. Yusa K, Rad R, Takeda J, Bradley A (2009) Generation of transgene-free CrossRef Medline induced pluripotent mouse stem cells by the piggyBac transposon. Nat Rhee J, Mahfooz NS, Arregui C, Lilien J, Balsamo J, VanBerkum MF (2002) Methods 6:363–369. CrossRef Medline